基于NMF 的老电影音频背景噪声修复算法

doi:10.12066/j.issn.1007-2861.1940

上海大学学报(自然科学版) ›› 2017, Vol. 23 ›› Issue (3): 333-341.doi: 10.12066/j.issn.1007-2861.1940

基于NMF 的老电影音频背景噪声修复算法

张叶君, 杨卫英

上海大学上海电影学院, 上海 200072

收稿日期:2017-05-02 出版日期:2017-06-30 发布日期:2017-06-30
通讯作者: 杨卫英(1957—), 女, 副教授, 研究方向为数字媒体技术等. E-mail: yangweiying@staff.shu.edu.cn
作者简介:杨卫英(1957—), 女, 副教授, 研究方向为数字媒体技术等. E-mail: yangweiying@staff.shu.edu.cn
基金资助:
国家自然科学基金资助项目(61571282)

Reduction of background audio noise for historical films based on non-negative matrix factorization

ZHANG Yejun, YANG Weiying

Shanghai Film Academy, Shanghai University, Shanghai 200072, China

Received:2017-05-02 Online:2017-06-30 Published:2017-06-30

摘要/Abstract

摘要：

老电影音频资料经过长时间的存储会出现音频纯度低、存在噪声等问题. 利用非负矩阵分解(non-negative matrix factorization, NMF)算法对单声道音频中的背景噪声进行自动、快速检测和分离以去噪. 对非噪声和噪声信号分别建立相应的模型, 即前者使用正弦模型; 后者的模型通过对老电影中先验噪声信号进行训练得到, 然后使用一种条件受限的NMF 算法对音频中的背景噪声进行分离. 实验结果表明, 该算法在去噪效果上要优于直接滤波等去噪算法.

关键词: 盲源分离, 音频去噪, 噪声模型训练, 非负矩阵分解

Abstract:

Audio materials of numerous historical films suffer from low sound quality, noise and other problems after being archived for a long time. This paper proposes a method based on non-negative matrix factorization (NMF) to automatically detect and separate background noise in a single channel audio. Harmonic signals and noises are modeled and differentiated using a sinusoid model and a priori noise training model respectively. Background noise is separated from the input audio with a constrained NMF algorithm. Experiments show that the proposed denoising algorithm outperforms the current algorithms in the denoise plug-in.

Key words: audio denoising, blind source separation, noise model training, non-negative matrix factorization

张叶君, 杨卫英. 基于NMF 的老电影音频背景噪声修复算法[J]. 上海大学学报(自然科学版), 2017, 23(3): 333-341.

ZHANG Yejun, YANG Weiying. Reduction of background audio noise for historical films based on non-negative matrix factorization[J]. Journal of Shanghai University（Natural Science Edition）, 2017, 23(3): 333-341.

参考文献

[1] Rumsey F. Challenges in archiving and restoration [J]. Journal of the Audio Engineering Society, 2016, 64(1): 94-97.
[2] Napieralska J. Concept of film sound restoration by adapting to contemporary cinema theatre [C]// 138th Audio Engineering Society Convention. 2015: 1031-1039.
[3] Lukin A, Todd J. Suppression of musical noise artifacts in audio noise reduction by adaptive 2-D filtering [C]// 123rd Audio Engineering Society Convention 123. Audio Engineering Society.
2007.
[4] Grancharov V, Samuelsson J, Kleijn B. On causal algorithms for speech enhancement [J]. IEEE Transactions on Audio Speech & Language Processing, 2006, 14(3): 764-773.

[5] Belouchrani A, Abed-Meraim K, Cardoso J F, et al. A blind source separation techinique based on second order statistics [J]. IEEE Transactions on Signal Processing, 1997, 45(2): 434-444.
[6] Lee D D, Seung H S. Learning the parts of objects by non-negative matrix factorization [J]. Nature, 1999, 401(6755): 788-791.
[7] Gillis N. The why and how of nonnegative matrix factorization [J]. Regularization, Optimization, Kernels, and Support Vector Machines, 2014, 12: 257-291.
[8] Weninger F, Roux J L, Hershey J R, et al. Discriminative NMF and its application to single-channel source separation [C]//15th Annual Conference of the International Speech Communication Association. 2014: 865-869.
[9] Zhou J, Chen S, Duan Z. Rotational reset strategy for online semi-supervised NMF-based speech enhancement for long recordings [C]// Applications of Signal Processing to Audio and
Acoustics. 2015: 1-5.
[10] F´evotte C, Bertin N, Durrieu J. Nonnegative matrix factorization with the Itakura-Saito divergence: with application to music analysis [J]. Neural Computation, 2009, 21(3): 793-830.
[11] Wilson K W, Raj B, Smaragdis P, et al. Speech denoising using nonnegative matrix factorization with priors [C]// IEEE International Conference on Acoustics, Speech and Signal
Processing. 2008: 4029-4032.
[12] Papadopoulos P, Vaz C, Narayanan S S. Noise aware and combined noise models for speech denoising in unknown noise conditions [C]// 17th Annual Conference of the International Speech Communication Association. 2016: 2866-2869.
[13] Lee D D. Algorithms for nonnegative matrix factorization [J]. Advances in Neural Information Processing Systems, 2001, 13(6): 556-562.
[14] Serra X. Musical sound modeling with sinusoids plus noise [M]. Amsterdam: Royal Swets & Zeitlinger, 1997: 91-122.
[15] McAulay R, Quatieri T. Speech analysis/synthesis based on a sinusoidal representation [J]. IEEE Transactions on Acoustics Speech & Signal Processing, 1986, 34(4): 744-754.
[16] Wang Y X, Zhang Y J. Nonnegative matrix factorization: a comprehensive review [J]. IEEE Transactions on Knowledge & Data Engineering, 2013, 25(6): 1336-1353.
[17] Bertin N, Badeau R, Vincent E. Fast Bayesian NMF algorithms enforcing harmonicity and temporal continuity in polyphonic music transcription [C]// IEEE Workshop on Applications of
Signal Processing to Audio and Acoustics. 2009: 29-32.
[18] Kim C, Stern R M. Robust signal-to-noise ratio estimation based on waveform amplitude distribution analysis [C]// 9th Annual Conference of the International Speech Communication
Association. 2008: 2598-2601.

基于NMF 的老电影音频背景噪声修复算法

Reduction of background audio noise for historical films based on non-negative matrix factorization

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价